Overview
Meta has launched Voicebox, a text-to-speech tool that uses generative AI to produce a synthetic voice 20 times faster than existing models, with only two seconds of recording. Meta said its deepfake voices are of such quality that it is not releasing all the code behind Voicebox, while it has also created a detector to recognise when synthetic speech has been used. The company said Voicebox could be used to offer natural-sounding voices to virtual assistants and non-player characters in the metaverse, as well as helping visually impaired people to hear written messages from friends read out.